Search

Tools to predict pollinator activity at regional scales generally rely on land cover maps, combined with human-inferred mechanistic rules and/or expert knowledge. Recently, Giménez-García et al. (2023) showed that, using large pollinator datasets, different environmental variables, and machine learning models, those predictions can be enhanced but at the cost of losing model interpretability. Here, we complement this work by exploring the potential of using advanced machine learning techniques to directly infer wild-bee visitation rates across different biomes only from land cover maps and available pollinator data while maintaining a mechanistic interpretation. In particular, we assess the ability of convolutional neural networks (CNNs), which are deep learning models, to infer mechanistic rules able to predict pollinator habitat use. At a global scale, our CNNs achieved a rank correlation coefficient of 0.44 between predictions and observations of pollinator visitation rates, doubling that of the previous human-inferred mechanistic models presented in Giménez-García et al. (2023) (0.17). Most interestingly, we show that the predictions depend on both landscape composition and configuration variables, with prediction rules being more complex than those of traditional mechanistic processes. We also demonstrate how CNNs can improve the predictions of our previous data-driven models that did not use land cover maps by creating a new model that combined the predictions of our CNN with those of our best regression model based on environmental variables, a Bayesian ridge regressor. This new ensemble model improved the overall rank correlation from 0.56 to 0.64.

Ecological intensification has been embraced with great interest by the academic sector but is still rarely taken up by farmers because monitoring the state of different ecological functions is not straightforward. Modelling tools can represent a more accessible alternative of measuring ecological functions, which could help promote their use amongst farmers and other decision-makers. In the case of crop pollination, modelling has traditionally followed either a mechanistic or a data-driven approach. Mechanistic models simulate the habitat preferences and foraging behaviour of pollinators, while data-driven models associate georeferenced variables with real observations. Here, we test these two approaches to predict pollination supply and validate these predictions using data from a newly released global dataset on pollinator visitation rates to different crops. We use one of the most extensively used models for the mechanistic approach, while for the data-driven approach, we select from among a comprehensive set of state-of-the-art machine-learning models. Moreover, we explore a mixed approach, where data-derived inputs, rather than expert assessment, inform the mechanistic model. We find that, at a global scale, machine-learning models work best, offering a rank correlation coefficient between predictions and observations of pollinator visitation rates of 0.56. In turn, the mechanistic model works moderately well at a global scale for wild bees other than bumblebees. Biomes characterized by temperate or Mediterranean forests show a better agreement between mechanistic model predictions and observations, probably due to more comprehensive ecological knowledge and therefore better parameterization of input variables for these biomes. This study highlights the challenges of transferring input variables across multiple biomes, as expected given the different composition of species in different biomes. Our results provide clear guidance on which pollination supply models perform best at different spatial scales – the first step towards bridging the stakeholder–academia gap in modelling ecosystem service delivery under ecological intensification.

Los avances tecnológicos y metodológicos de las últimas décadas (e.g., información satelital, potencia de los ordenadores, análisis geoespacial, desarrollo de algoritmos) facilitaron buscar soluciones a problemas complejos como el cambio global. Estos avances permitieron que surjan plataformas informáticas para modelar servicios ecosistémicos, que cuantifican los beneficios de la naturaleza y evalúan cómo son o serán afectados por acciones humanas. Actualmente, existen variadas plataformas con diferentes grados de aptitud según el contexto, destacándose k.LAB por ser gratuita, de código abierto y presentar un enfoque de ciencia colaborativa, además de integrar diferentes técnicas de modelado con inteligencia artificial. k.LAB es muy versátil para responder a las demandas de diferentes usos, desde programar y modelar SE hasta tomar decisiones. Sin embargo, quienes cuantifican y mapean SE, especialmente en Latinoamérica, tienen escaso conocimiento de k.LAB; esto dificulta aprovechar su potencial, tal como sucedió con herramientas de acceso libre y código abierto (e.g., la adopción de R requirió tiempo, revisiones, discusiones y materiales didácticos en revistas especializadas). Este trabajo presenta las capacidades de k.LAB en el contexto de las plataformas de modelado de SE. Primero, introducimos estas plataformas en términos generales, con énfasis en las más usadas. Luego, caracterizamos k.LAB técnica y filosóficamente. Después, presentamos un caso de estudio en el norte de la Patagonia argentina, ilustrando la obtención de mapas de tres SE (captura de carbono, polinización y recreación al aire libre) utilizando aplicaciones de modelado dirigidas a personas sin experiencia en programación. Finalmente, establecemos características deseables en las plataformas de modelado de SE para discutir ventajas y limitaciones de k.LAB en relación con otras alternativas. Esperamos brindar un marco general útil para el modelado de SE y ampliar el conjunto de herramientas para abordar problemáticas vinculadas al cambio global en la Argentina y otros países de la región.

Predicting how communities re-arrange in response to changes in species composition remains a key challenge in ecology. Migratory species, which enter and leave communities across latitudinal gradients, offer us a unique opportunity to evaluate community- and species-level responses to a shift in community composition. We focused on a migratory hummingbird and the communities that host it along a latitudinal and species diversity gradient. Our results show higher niche overlap in more diverse communities, allowing resident species to compensate for the loss of the migrant in providing pollination services. Contrastingly, in less diverse communities, the migrant behaves as a specialist, monopolizing abundant resources. In its absence, its role is not fully covered by resident species, resulting in a decrease in the fruit set of the migrant's preferred plant species. These results help us understand the potential impacts of biodiversity loss and have important implications for community persistence given expected changes in the migratory behaviours of some species.

Abstract In recent years, an extended body of literature has focused on the importance of either temporal or spatial dynamics in shaping the structure of interacting plant and pollinator communities. This improvement from a previously static and aggregated perspective has allowed us to understand many of the ecological processes that shape community assembly. However, fewer are the studies that have simultaneously focused on spatial and temporal dynamics, and even fewer are those that collect data across different habitat types to assess the generality of their findings. Here, we used a dataset collected weekly throughout the full flowering season for two consecutive years and within two contrasting habitat types in N and SW Spain: a mountain grassland area and the understory of sparse pine forests. We evaluated species and interaction persistence through space and time, pollinator fidelity, and turnover patterns in interaction composition while providing a potential mechanistic explanation for the patterns observed. Our results show that although species generalization does not explain species or interaction persistence, moderately generalist species are those showing the greatest fidelity to the subset of plant species they visit through space and time. Further, we find that interaction turnover through time is mostly driven by changes in species composition, while through space it is mostly driven by interaction rewiring resulting from indirect competitive interactions between pollinator species. Our results help to shed light on the potential mechanisms driving community assembly patterns beyond niche or neutral processes by adding within‐trophic‐level interactions that can modify pollinator preferences.

Abstract Progress in key social-ecological challenges of the global environmental agenda (e.g., climate change, biodiversity conservation, Sustainable Development Goals) is hampered by a lack of integration and synthesis of existing scientific evidence. Facing a fast-increasing volume of data, information remains compartmentalized to pre-defined scales and fields, rarely building its way up to collective knowledge. Today's distributed corpus of human intelligence, including the scientific publication system, cannot be exploited with the efficiency needed to meet current evidence synthesis challenges; computer-based intelligence could assist this task. Artificial Intelligence (AI)-based approaches underlain by semantics and machine reasoning offer a constructive way forward, but depend on greater understanding of these technologies by the science and policy communities and coordination of their use. By labelling web-based scientific information to become readable by both humans and computers, machines can search, organize, reuse, combine and synthesize information quickly and in novel ways. Modern open science infrastructure—i.e., public data and model repositories—is a useful starting point, but without shared semantics and common standards for machine actionable data and models, our collective ability to build, grow, and share a collective knowledge base will remain limited. The application of semantic and machine reasoning technologies by a broad community of scientists and decision makers will favour open synthesis to contribute and reuse knowledge and apply it toward decision making.

Abstract The search for healthy diets has led to a surge in the demand for functional foods or ‘superfoods’, which have now become popular among the middle‐ and high‐income fractions of the society in developed regions of the world. ‘Superfoods’ are predominantly consumed far from their centres of origin and out of their cultural context with different environmental and social effects. Here, we present a series of case studies to provide an overview of the different environmental impacts driven by superfood expansion. We show that if these crops are to follow the path of other global commodities, then strong environmental impacts and large carbon footprints are expected in terms of land clearing, use of agrochemicals and transportation during times of high prices (boom) and social problems as farmers have to abandon their livelihoods when prices sink below the cost of production (bust). We also showcase how a combination of management practices, consumer choices and policy changes could help in alleviating the ecological footprint of these crops. A free Plain Language Summary can be found within the Supporting Information of this article.

Our understanding of how the structure of species interactions shapes natural communities has increased, particularly regarding plant-pollinator interactions. However, research linking pollinator diversity to reproductive success has focused on pairwise plant-pollinator interactions, largely overlooking community-level dynamics. Here, we present one of the first empirical studies linking pollinator visitation to plant reproduction from a community-wide perspective. We use a well-replicated dataset encompassing 16 plant-pollinator networks and data on reproductive success for 19 plant species from Mediterranean shrub ecosystems. We find that statistical models including simple visitation metrics are sufficient to explain the variability observed. However, a mechanistic understanding of how pollinator diversity affects reproductive success requires additional information on network structure. Specifically, we find positive effects of increasing complementarity in the plant species visited by different pollinators on plant reproductive success. Hence, maintaining communities with a diversity of species but also of functions is paramount to preserving plant diversity.

Forest fragmentation is a grave threat to biodiversity. Forests are becoming increasingly fragmented with more than 70% now < 1 km from forest edge. Although much is known about the effects of forest fragmentation on individual species, much less is understood about its effects on species interactions (i.e., mutualisms, antagonisms, etc.). In 2014, a previous meta‐analysis assessed the impacts of forest fragmentation on different species interactions, across 82 studies. We pooled the previous data with data published in the last 10 years (combined total 104 studies and 168 effect sizes). We compared the new set of publications (22 studies and 32 effect sizes) with the old set to evaluate potential changes in species interactions over time given the global increase in fragmentation rates. Mutualisms were more negatively affected by forest fragmentation than antagonisms (p < 0.0001). Edge effects, fragment size, and degradation negatively affected mutualisms, but not antagonisms, a different finding from the original meta‐analysis. Parasitic interactions increased as fragment size decreased (p < 0.0001)—an intriguing result at variance with earlier studies. New publications showed a more negative mean effect size of forest fragmentation on mutualisms than old publications. Although research is still limited for some interactions, we identified an important scientific trend: current research tends to focus on antagonisms. We concluded that forest fragmentation disrupts important species interactions and that this disruption has increased over time. [ABSTRACT FROM AUTHOR]

Abstract. The study of ecological stability continues to fill the pages of scientific journals almost seven decades after the first ecologists initiated this line of research. The many advances in this field have focused on understanding the stability of populations, communities or functions within single guilds or trophic levels, with less research conducted across multiple trophic levels and considering the different interactions that relate species to each other. Here, we review the recent literature on the multiple dimensions of ecological stability specifically within plant–pollinator communities. We then focus on one of stability´s dimensions, temporal invariability, and adapt an existing partitioning framework that bridges invariability and synchrony measures across spatial scales and organizational levels to accommodate interactions between plants and their pollinators. Finally, we use this framework to analyse temporal invariability in plant reproductive success, partitioning it on invariability and synchrony components across plant and pollinator populations and communities, as well as their interactions, using a well-resolved dataset that encompasses data for two years. Our review of the literature points to several significant gaps in our current knowledge, with simulation studies clearly overrepresented in the literature as opposed to experimental or empirical approaches. Our quantitative approach to partitioning invariability shows similar patterns of decreasing temporal invariability across increasing organizational levels driven by asynchronous dynamics amongst populations and communities, which overall stabilize ecosystem functioning (plant reproductive success). This study represents a first step towards a better comprehension of temporal invariability in ecosystem functions defined by interactions between species and provides a blueprint for the type of spatially replicated multi-year data that needs to be collected in the future to further our understanding of ecological stability within multi-trophic communities. [ABSTRACT FROM AUTHOR]

Abstract Human-induced forest fragmentation poses one of the largest threats to global diversity yet its impact on rattans (climbing palms) has remained virtually unexplored. Rattan is arguably the world’s most valuable non-timber forest product though current levels of harvesting and land-use change place wild populations at risk. To assess rattan response to fragmentation exclusive of harvesting impacts we examined rattan abundance, demography and ecology within the forests of northeastern, Australia. We assessed the community abundance of rattans, and component adult (>3 m) and juvenile (≤3 m) abundance in five intact forests and five fragments (23–58 ha) to determine their response to a range of environmental and ecological parameters. Fragmented forests supported higher abundances of rattans than intact forests. Fragment size and edge degradation significantly increased adult rattan abundance, with more in smaller fragments and near edges. Our findings suggest that rattan increase within fragments is due to canopy disturbance of forest edges resulting in preferential, high-light habitat. However, adult and juvenile rattans may respond inconsistently to fragmentation. In managed forest fragments, a rattan abundance increase may provide economic benefits through sustainable harvesting practices. However, rattan increases in protected area forest fragments could negatively impact conservation outcomes.

Habitat loss and urbanization are primary components of human impact on the environment. Here, Venter et al.use global data on infrastructure, agriculture, and urbanization to show that the human footprint is growing slower than the human population, but footprints are increasing in biodiverse regions.